Construction of Kansei engineering interface with respect to design and manufacturing of food suitable for consumer fevorableness

为了设计和制造出适合消费者胃口的食品，研究了感性工学界面的构建。(1)建立质量模型，从咖啡豆的混合比例预测感官评价分数。用三种有代表性的咖啡豆制备了22种混合咖啡，并由专家小组和使用响应面法（RSM）、多元回归分析（MRA）和模糊神经网络（FNN）构建的模型对10种感官属性进行了评估。RSM和MRA模型在某些感官属性上表现出良好的相关性，但总体准确性不足。FNN模型对所有属性具有较高的相关性，清晰地展示了混合比例与风味特性之间的关系，具有足够的准确性，可用于实际应用。因此，它构成了加速产品开发的有力工具。(2)为了确定工艺变量，从上到下进行了逆向计算。在这种情况下，遗传算法（GA）作为一种快速方便的搜索方法经常被使用。但是，如果与涉及数据的空间宽度相比，学习数据相对较少，则搜索解永远没有置信度，与正确解完全不同。为了克服这个问题，提出了伴随置信度估计的遗传算法（CFGA）。选取20多个方程作为机密函数（CF）的候选方程，以估计每个解的置信度。当最接近的三个数据点的误差和这些欧几里德距离都定义置信度时，所提出的遗传算法的搜索结果的正确性变得很高。此外，提出了基于CF的主动学习方法用于FNN建模。使用CF，我们可以知道在定位空间中有多少数据点。因此，可以主动提出数据点所需要的定位空间。在计算实验中，对一些数学方程进行了验证。作为一个模型空间。CF被认为是一种有效的辅助主动学习方法。(3)将CFGA应用于咖啡混合比例的确定和酒曲捣碎过程工艺变量的确定。利用CFGA对混合比例和工艺变量进行了估计，具有较高的准确性。(4)为了开发一个任何人都能方便使用的FNN建模软件，对FNN进行了封装。实现了软件的原型，包括输入变量选择的参数递增法、感性工程的核心程序FNN建模和逆向计算的遗传算法。少

In order to design and manufacture the food suitable for consumer fevorableness, construction of Kansei engineering interface was investigated.(1) Quality models were constructed to predict sensory evaluation scores from the blending ratio of coffee beans. Twenty-two blended coffees were prepared from three representative beans and were evaluated with respect to ten sensory attributes by an expert panel and by models constructed using the response surface method (RSM), multiple regression analysis (MRA), and a fuzzy neural network (FNN). The RSM and MRA models showed good correlations for some sensory attributes, but lacked sufficient overall accuracy. The FNN model exhibited high correlations for all attributes, clearly demonstrated the relationships between blending ratio and flavor characteristics, and was accurate enough for practical use. It thus constitutes a powerful tool for accelerating product development.(2) In order to determine process variable, reverse calculation from fo … More od design was investigated. In such cases, genetic algorithm (GA) has been often used as a speedy and convenient searching method. However, if the learning data is relatively fewer compared with the width of the space involving the data, searched solution has never the confidence and it becomes completely different with correct solution. To overcome this problem, GA accompanied with estimation of confidence (CFGA) was proposed. More than 20 equations were selected as a candidate of confidential function (CF) in order to estimate the confidence of each solution. When the confidence was defined by both of errors of the nearest three data points and those Euclidian distances, correctness of searching results by the proposed GA became high.In addition, active learning method using CF was proposed for FNN modeling. Using CF, we can know how much are there a crowd of data point in the located space. Therefore, the located space that the data points are needed can be actively suggested. In the calculation experiment, some mathematical equation was tested. as a model space. CF was found to be effective as a supporting method of active learning.(3) CFGA was applied for determination of coffee blending ratio and determination of process variables of Koji mashing process. Blending ratio and process variables were estimated with high correctness by the use of CFGA.(4) In order to develop a software of FNN modeling, of which the use can be easy for any person, FNN packaging was carried out. The prototype of software was achieved and it was including parameter increasing method for selection of input variables, FNN modeling as a core program for Kansei engineering, and GA for reverse calculation. Less